tkgally

17 ms·

1. 241.▲
   by tkgally 2y ago

   Thanks! Feel free to link to the HN comment. That will encourage me to make a video or two for YouTube demonstrating for a wider audience how I use AI for translation. I hope to do that within a few weeks.
2. 242.▲
   by tkgally 2y ago

   I don’t know. I stopped using DeepL sometime last year as I found its inability to be prompted about the purpose of the translation to be too limiting for my purposes. At that time, it also had problems with things like maintaining coherent
3. 243.▲
   by tkgally 2y ago

   There might be some papers or other guides out there, but their advice will be based on whatever tools happened to be available at the time they were written and on the particular types of translations the authors cared about. The technolog
4. 244.▲
   by tkgally 2y ago

   That might work to prevent the glitches I noticed toward the end of the long texts. In general, though, I haven’t seen any sign yet that reasoning models are better at translation than nonreasoning ones even with shorter texts. LLMs in gene
5. 245.▲
   by tkgally 2y ago

   As I explained in a sister comment, it is not possible to rate translation quality objectively, as opinions and positions about what constitutes a good translation vary. But in my tests of reasoning models since the release of o1-preview, t
6. 246.▲
   by tkgally 2y ago

   No. People’s tastes and judgments vary too much. One fundamental area of disagreement is how closely a translation should reflect the content and structure of the original text versus how smooth and natural it should sound in the target lan
7. 247.▲
   by tkgally 2y ago

   At the end of his post, Simon mentions translation between human languages. While maybe not directly related to token limits, I just did a test in which both R1 and o3-mini got worse at translation in the latter half of a long text. I ran t
8. 248.▲
   by tkgally 2y ago

   Borrowing most of Simon’s prompt, I tried the following with o3-mini-high in the chat interface with Search turned on: “Summarize the themes of the opinions expressed in discussions on Hacker News on January 31 and February 1, 2025, about O
9. 249.▲
   by tkgally 2y ago

   Many previous HN discussions: https://hn.algolia.com/?q=sabotage+manual
10. 250.▲
    by tkgally 2y ago

    I’ve been reading opinion columnists at the New York Times for nearly a half century. My opinion is that the paper should stop giving de-facto lifetime tenure to columnists. All of the long-term columnists I remember—with the possible excep
11. 251.▲
    by tkgally 2y ago

    Twice last year, I went into a coffee shop in Osaka to kill some time and the shop turned out to allow smoking, so I left. I was surprised both times, as that hasn’t happened to me in Yokohama (where I live) or Tokyo for a long time. One of
12. 252.▲
    by tkgally 2y ago

    Minor correction: GPT-4 was announced on March 14, 2023, less than two years ago. I don’t remember how much LLMs had been discussed as coding assistants before then, but it was Greg Brockman’s demonstration of using it to write code that fi
13. 253.▲
    by tkgally 2y ago

    There have been discussions on HN before about political issues that are censored by Chinese LLMs. This video covers some of those, but he also gives examples of technical and coding-related topics that are blocked by DeepSeek.
14. 254.▲
    DeepSeek has a major issue [video](youtube.com)

    1 points by tkgally 2y ago|2 comments
15. 255.▲
    by tkgally 2y ago

    I agree that many aspects of intelligence—and of the lack of intelligence—are not being measured by such benchmarks. One issue is they are only examining problems that have right answers. One of the most powerful uses of LLMs for me, at lea
16. 256.▲
    by tkgally 2y ago

    Thanks for trying! I was thinking of creating a set of digits for base 26 using emoji or other Unicode characters. Then the numbers would at least look like palindromes. But I don't know how interesting palindromes in an ad hoc charact
17. 257.▲
    by tkgally 2y ago

    How about representing the words as numbers in base 26? That would preserve the palidromicity for both the word and the number, and primality would not be affected by the base representation. Might not be as much fun, though.
18. 258.▲
    by tkgally 2y ago

    Interesting! What do you do when you try to factor a large number that doesn't yield to any of your mental factoring techniques? It might be prime, after all, or composite with only large prime factors. Do you give up after a while? Ru
19. 259.▲
    by tkgally 2y ago

    Follow-up: My formal study of programming stopped with that Fortran IV class a half century ago, but LLMs now delude me into thinking I can program. I just had Claude write me a Python program to list all of the seven-digit numbers that are
20. 260.▲
    by tkgally 2y ago

    > In 1985, she published a paper titled “How Often Is the Number of Divisors of n a Divisor of n?” Claudia Spiro seems to have remained actively interested in prime numbers into her sixties. In 2017, she published a paper titled “On thre
21. 261.▲
    by tkgally 2y ago

    I haven’t heard of any. It would great if there were....
22. 262.▲
    Anthropic's Race to Build a Smarter Claude and Human-Level AI [video](youtube.com)

    4 points by tkgally 2y ago|0 comments
23. 263.▲
    by tkgally 2y ago

    My family’s phone number when I was a child was both a palindrome and a prime: 7984897. My parents had had the number for two decades without noticing it was a palindrome. I still remember my father’s delight when he got off a phone call wi
24. 264.▲
    by tkgally 2y ago

    What models have you been using for that? While I haven’t tried automating the production of vocabulary lists through an API, within the last few weeks I have had the chat versions of ChatGPT 4o, Claude Sonnet 3.5, and one of the latest Gem
25. 265.▲
    by tkgally 2y ago

    This looks really good. I wish I had had something like this many years ago when I was studying languages. Somebody has already suggested adding spaced repetition and audio, which I agree with completely. One more suggestion: In addition to
26. 266.▲
    by tkgally 2y ago

    I use it at https://chat.deepseek.com/ . It’s free but requires a log-in. Now, when I hover over the “DeepThink” button below the prompt field, a pop-up appears saying “Use DeepSeek-R1 to solve reasoning problems.”
27. 267.▲
    by tkgally 2y ago

    Over the last two weeks, I ran several unsystematic comparisons of three reasoning models: ChatGPT o1, DeepSeek’s then-current DeepThink, and Gemini 2.0 Flash Thinking Experimental. My tests involved natural-language problems: grammatical a
28. 268.▲
    by tkgally 2y ago

    Thank you! The book is out of date and better language-learning tools are available now, but if anyone might find an old-fashioned phrasebook/textbook useful or interesting, they can download the PDF for free, no strings attached: htt
29. 269.▲
    by tkgally 2y ago

    Thanks!
30. 270.▲
    by tkgally 2y ago

    Nice article. I have only one chopstick sleeve in my collection, a humorous, slightly risqué one from a ramen and gyoza shop in Yokohama that I frequent. Here is a photograph with transcription and translation: https://gally.net&